Hi all,
I am struggling with some conceptual problems concerning the Annex-H proposal of Edgar Martinez, feeling that this has some commonality with the remarks of Roberto Winkler and Simon Binar on it. IMHO, the proposed architecture, and especially the WAU and the GK, are mixing all kinds of mobility related functionality in a single, strongly coupled, do-it-all-in-H.323 layer, involving radio access up to H.323 application layer.
Specifically, I have some difficulties with the following:
1. No clear division in terminal and user mobility aspects in the basic H.323 architecture is taken as a starting point: shouldn't there be specified somewhere in general, in conjunction with the H.323 set of protocols, that IP addresses represent terminals in an IP network, and that users can only be represented by logical names from a directory (i.e. aliases), and whether an "endpoint" is a user or a terminal? If this is decided on, the H.323 registration procedure is just what is needed to have user mobility, by dynamically associating a logical user name with one or multiple terminals upon user request. With terminals identified by IP addresses, terminal mobility would further be strictly a network access problem.
2. Annex-H is supposed to handle user and service mobility, so why would the handover problem, as any other terminal mobility related problem, be handled here instead of independently in Annex-I?
3. Mobility is not an issue in H.323 alone, e.g. consider roaming and handover for a connection-oriented data application, so why should it be handled in H.323 on its own? Or put otherwise, is H.323 support really required on a terminal in order to be able to roam, as a user of some service, or as a terminal accessing a network?
4. Although the proposal pays attention to the identification of several degrees of locality for roaming and handover, solving all control issues (radio access, IP-networking, etc.) in a single application layer still leads to both resource sub-optimality and complexity. E.g. why should the WAU be aware of anything having to do with (user) mobility management?
Instead I would propose to use an approach of highly independent layers, following good Internet tradition, and solving the problems as locally as possible (both geographically and architecturally). Here's the layers I think at least should be distinguished:
- MAC level: terminal micro-mobility across radio cells, ethernet segments, etc. can exist without any consciousness of this at the IP level (e.g. using wireless LAN tunneling techniques);
- IP network level: terminal macro-mobility across the Internet should be invisible to higher layers (e.g. by means of Mobile IP v4 or v6 and the improvements suggested for delay-sensitive traffic like RTP streams: Mobile IP HAWAII, or draft-elmalki-mobileip-fast-handoffs-01 to name but a few; note that route optimization is also worked on in this area);
- "terminal real-time signaling" application level: the level where telephony and MM signaling like H.323 can be placed (restricting GW and GK functionality to strictly this level); here Annex-H might define transport for additional parameters and vertical signaling, useful for serving the "user services" application level (below); at this lowest application level packet to circuit switched network boundaries may be crossed (cf. the evolving traditional telephony networks vs. the Internet);
- "user services" application level: user and service mobility takes place here, making use of a network- (and H.323) independent naming system (directories using e.g. DNS, LDAP schemas, etc.) and e.g. back-end services, "behind" or "on top of" one or several GKs for realizing UM, user mobility support for a larger scope than just H.323, smart user assistance, etc..
This approach has not only the general advantage that every layer can evolve technically independent of all others, it also has the specific advantage that it realizes true consolidation of mobility for telephony and traditional data applications.
As a last point, it is not clear to me at which of these levels H.450 would belong. As it is specified now it is on top of H.323, but suppose one would like to adopt a service provision model where a user could subscribe to a telephony-related service , wouldn't it be nice then to be able to offer this across different telephony technologies (a bit like where IN aims at)? A comparison to using a HTTP service control layer (Annex-K) might be interesting here.
I cannot imagine that these fundamental issues where not previously discussed in ITU-T SG16 or Tiphon WG7, so if anyone can give me some arguments why an approach as I tried to describe here could not be followed, I would be glad to know about it.
Regards, Marc Roelands Siemens ICN Atea Atealaan 34 B-2200 Herentals Belgium Tel.: +32 14253965 Fax: +32 14222994 E-mail: marc.roelands@siemens.atea.be
-----Original Message----- From: Roberto Winkler [mailto:wnk@FUB.IT] Sent: Tuesday, September 14, 1999 10:34 AM To: TIPHON_WG7@LIST.ETSI.FR Subject: Re: [Fwd: Re: [Fwd: [Fwd: [Fwd: Re: H.323 mobility first darft]]]]
Jin, Edgar,
If I may try to give my 2 cents in this discussion, I think that the problems/weakness of 3GPP considered systems (which Edgar is speaking about) cannot match with the faults identified by Lucent's contribution, which is focused on the unique issue of R99 terminal support in the All IP R00 network. Considering these 3GPP contributions does not help in understanding Edgar's point of view. I have already sent a few e-mails to comment the proposed mobile H.323 architecture and would like to summarize here my view, hoping to get reactions: - the role and reason for being of WAU is unclear (as TIPHON is focused on user and service mobility, why worry about the access medium ?) - the redefinition of several RAS channel messages is confusing (why does the RAS channel terminates at WAU ?) - protocol stacks and interface definitions are missing from the picture.
best regards